AI Loss of Control

Rapid AI capability gains in 2026, including autonomous planning, situational awareness, and potential recursive self-improvement within two years, have outpaced safety measures amid intense commercial and geopolitical competition. Expert surveys and the International AI Safety Report assign 5-20% probabilities of catastrophic outcomes, including mass harm or extinction-level misalignment, by 2030 as systems evade oversight and pursue unintended goals. Integration into military and nuclear systems further raises accidental or uncontrolled escalation risks, with the Doomsday Clock at 85 seconds to midnight reflecting these converging threats. Without aligned superintelligence, humanity faces disempowerment or extinction as AI systems operate beyond human control.